Examining Relationships between Leisure Involvement ... · loyalty, stating that "the concept of commitment provides an essential basis for distinguishing between brand loyalty and

Journal of Leisure Research Copyright 20042004, Vol. 36, No. 1, pp. 45-72 National Recreation and Park Association

Examining Relationships between LeisureInvolvement, Psychological Commitment and

Loyalty to a Recreation Agency

Yoshi Iwasaki, Ph.D.University of Manitoba

Mark E. Havitz, Ph.D.University of Waterloo

Despite the recent proliferation of leisure involvement and loyalty research, verylittle attention has been given to systematically conceptualize and examine thenature of leisure involvement's relationship with loyalty. The purpose of thepresent study was to examine whether psychological commitment intervenes inthe relationship between fitness participants' leisure involvement and their be-havioral loyalty to a recreation agency. Concomitant consideration was alsogiven to the moderating effects of selected personal and social factors on thisrelationship (Iwasaki & Havitz, 1998). Comparisons of a Fully Mediated Model(F-M-M) with rival Direct Effects Models (D-E-M I & II), using structural equa-tion modeling (SEM), suggested that psychological commitment mediated theeffects of enduring involvement on patrons' behavioral loyalty (p < .01). Thatis, enduring involvement appeared to indirectly influence behavioral loyalty viaa sequence of commitment factors. Significant evidence was also found for thedirect effects of skill, motivation, social support, and social norms on enduringinvolvement, while some support was found for moderating effects (i.e., skill,motivation, social support, and side bets significantly moderated the effects ofenduring involvement on commitment's formative factors). Overall, the find-ings of the study help uncover some of the key mechanisms/processes by whichcustomers become loyal to a recreation agency. Marketing strategies may bedeveloped to aim at strengthening customer loyalty by maximizing key ante-cedents of behavioral loyalty (i.e., enduring involvement and psychological com-mitment), as well as personal and social factors (e.g., skill, motivation, socialsupport, social norms, and side bets/sunk costs).

KEYWORDS: Enduring involvement, commitment, loyalty, recreation services, leisurebehavior.

Research on leisure involvement and loyalty has important implications.Recreation service agencies can benefit from having involved and loyal par-

Address correspondence to: Yoshi Iwasaki, 102 Frank Kennedy Centre, Winnipeg, Manitoba,Canada R3T 2N2. Phone: (204) 474-8643; Fax: (204) 474-7634. Email: [email protected].

Author Note: Yoshi Iwasaki is in the Health, Leisure & Human Performance Research In-stitute, Faculty of Physical Education and Recreation Studies, at University of Manitoba. MarkHavitz is in the Department of Recreation and Leisure Studies at University of Waterloo. Thisresearch was supported by the University of Manitoba/University Research Grants Program. Wegratefully acknowledge the insightful comments from the JLR reviewers which substantially im-proved the quality of the final manuscript. Any errors are, of course, our sole responsibility.

45

4 6 IWASAKI AND HAVTTZ

ticipants/clients since these participants or clients appear to play an essentialrole in achieving important organizational goals, including revenue gener-ation, developing a positive reputation, networking, community develop-ment, and promoting the quality of life (Backman & Crompton, 1991; Bul-laro & Edginton, 1986; Gahwiler & Havitz, 1998; Howard, Edginton, & Selin,1988). Thus, developing and maintaining customer loyalty brings "a sustain-able competitive advantage" (Dick & Basu, 1994, p. 99) and has become "astrategic mandate in today's service markets" (Ganesh, Arnold, & Reynolds,2000, p. 65).

Leisure involvement and loyalty have implications for ongoing partici-pation and client retention. For example, Gahwiler (1995) pointed out thatannual retention rates of membership-based fitness facilities are, on average,only 50%. Howard (1992) found that only 2% of American adults accountedfor 75% of annual participation in six leisure activities examined, includinggolf and running. More recently, Barber and Havitz (2001) examined adultCanadians' participation rates in ten sport and fitness activities. Similar toHoward (1992), Barber and Havitz split activity individuals into occasional,regular, and avid participants. They reported that avid participants accountedfor a low of 70% of all bicycling to a high of 84% of all running/joggingand ice hockey participation. Furthermore, research evidence suggests thatit is more efficient (i.e., up to six times) for practitioners to focus on retain-ing current clients than on seeking new ones (O'Boyle, 1983).

It has been shown that leisure involvement and loyalty are related butdistinct constructs (e.g., Backman & Crompton, 1991; Park, 1996), whilepsychological commitment has been conceptualized as a key linking variablebetween these two constructs (Iwasaki & Havitz, 1998). Briefly, leisure in-volvement reflects people's beliefs about their leisure participation includingthe importance of and interest in such participation, and symbolic valuesderived from it (Havitz & Dimanche, 1997). Psychological commitment rep-resents people's attitude toward a brand (e.g., a recreation service provider)such as their resistance to change their preferences toward the brand (Prit-chard, Havitz, & Howard, 1999). Finally, loyalty reflects people's attitude andbehavior toward a brand of service and repeat patronage in the use of thebrand (Backman & Crompton, 1991; Park, 1996).

In contemporary leisure research, leisure involvement and loyalty havebeen two of the widely studied topics. The dimensions of leisure involvementhave been extensively examined, along with their consequences (e.g., pur-chase decisions, participation patterns, and recreation service promotion).Other issues examined include the temporal stability of leisure involvementand its associations with user characteristics (see Havitz & Dimanche, 1997,1999 for reviews of leisure involvement research). Loyalty in leisure settingshas been discussed in both activity and service provider contexts, with re-searchers defining the phenomenon using both behavioral and attitudinalindicators. Participants and clients have been segmented based on their typesand levels of loyalty, and antecedents of loyalty have been examined (e.g.,Backman & Crompton, 1991; Howard et al., 1988; Pritchard et al., 1999;Pritchard, Howard, & Havitz, 1992; Selin, Howard, Udd, & Cable, 1988).

INVOLVEMENT, COMMITMENT AND LOYALTY 47

Also, conceptual links between loyalty and other constructs such as con-straints and service quality have been explored (e.g., Backman, 1991; Back-man & Veldkamp, 1995).

Generally, researchers have independently examined the concepts ofleisure involvement and loyalty. Very little attention has been given to system-atically conceptualize and/or directly examine the potential relationshipsbetween the two concepts. Although Backman and Crompton (1991) andPark (1996) reported correlations that suggest these two concepts are dis-tinct, the detailed process by which leisure involvement influences clientloyalty is largely unexplained. In an attempt to overcome this limitation,Iwasaki and Havitz (1998) developed a conceptual model to explain the com-plexity of leisure involvement's relationships with participant and clientloyalty.

Relationships between Leisure Involvement and Behavioral Loyalty

Iwasaki and Havitz's (1998) theoretical model proposed relationshipsbetween individuals' involvement in a leisure activity and their behavioralloyalty to a recreation agency. Figure 1 represents their revised model to

- •- •

Personal

Antecedents(e.g., values, attitudes,

motivation, needs,skills, intrapersonal

constraints)

Involvement

AttractionSign

CentralityRisk Probability

Risk Consequence

i

Social-

Situational

Antecedents(e.g., social support,

social norms, structuralconstraints)

/

/

\

\

f

Personal Moderators(e.g., attitude accessibility, emotions, personal benefits, side bets, switching costs, skills)

Commitment's

Formative Factors

InformationalComplexity

Cognitive ConsistencyConfidence

Position InvolvementVolitional Choice

11

Psychological

Commitment

Resistance

to Change >

Behavioral

Loyalty

DurationFrequencyIntensitySequenceProportionProbability

Social-Situational Moderators(e.g., social support, situational incentives, social-cultural norms,

interpersonal or structural constraints)

=

Note: A revised version of the model adopted originally from Iwasaki and Havitz (1998)

Figure 1. A Conceptual Model of the Relationships among Involvement, PsychologicalCommitment, and Behavioral Loyalty

4 8 IWASAKI AND HAVITZ

more appropriately reflect current conceptualizations. The revised model1

argues that becoming a loyal client is entailed by becoming highly involvedin a leisure activity and developing psychological commitment to a recreationagency. Psychological commitment is comprised of formative factors (e.g.,informational complexity, cognitive consistency) and attitudes toward resis-tance to change agency preferences. The model also posits that processesleading to loyalty differ according to personal characteristics (e.g., motiva-tion, skills) and social-situational factors (e.g., social support, social norms).In Figure 1, dotted lines represent interaction effects (i.e., moderating ef-fects) between different constructs (e.g., skills X involvement, social supportX resistance to change) for illustration purposes. Thus, the model consistsof two major components: (a) the mediating role of psychological commit-ment in the relationships between leisure involvement and behavioral loyalty(mediating component), and (b) the moderating effects of personal and social-situational factors on such relationships (moderating component).

Iwasaki and Havitz's (1998) model is derived from existing conceptualframeworks. Day (1969) argued that true loyalty, as opposed to repeat pur-chasing, exists only when consumers are highly involved with their purchasesand highly committed to a particular brand. Beatty, Kahle, and Homer(1988) developed an involvement-commitment model outlining how endur-ing or ego involvement leads to brand commitment and found empiricalsupport for their model. Similarly, Crosby and Taylor (1983) stated that "in-volvement often precedes commitment" (p. 415). On the other end, Jacobyand Chesnut (1978) described the connection between commitment andloyalty, stating that "the concept of commitment provides an essential basisfor distinguishing between brand loyalty and other forms of repeat purchas-ing behavior" (p. 84). Kelley and Davis (1994) and Morgan and Hunt (1994)further clarified the role of commitment by finding the mediating role itplayed in key outcomes of consumer or purchasing behavior including loy-alty. Furthermore, Crosby and Taylor's (1983) definition of psychologicalcommitment emphasized "resistance to change" as the construct's root ten-dency of psychological commitment that develops from antecedent or for-mative factors (i.e., informational, identification, and volitional processes).Many of the antecedents linked to loyalty in Iwasaki and Havitz's (1998)model are consistent with Dick and Basu's (1994) theoretical framework ofattitudinal precursors to loyalty's formation.

Recently, Pritchard et al. (1999) developed a conceptual model of com-mitment and its link with loyalty (i.e., a Mediated Effects Model) to helpexplain an important question—how and why do clients develop a sense of

'Iwasaki and Havitz's (1998) original model illustrated that resistance to change was a discreteor separate outcome of psychological commitment. Instead, their revised model described heresuggests that psychological commitment is defined by its formative factors and resistance tochange—resistance to change is considered an integral part of or the evidence of psychologicalcommitment. The latter description is more consistent with the current conceptualizations ofpsychological commitment.


loyalty? They tested the validity of the Mediated Effects Model against itsrival models, Direct Effects Models, using a sample of airline and hotel pa-trons, and found evidence that clients' resistance to change is best consid-ered as an integral part of psychological commitment, and that it is a keyprecursor to the development of their loyalty.

In addition to being derived from the consumer behavior literature,Iwasaki and Havitz's (1998) model is grounded in the literature focusing onleisure and recreation. For example, Lee and Zeiss (1980) found that highlycommitted major league baseball spectators tend to show loyalty to theirfavorite teams. Siegenthaler and Lam's (1992) analyses of involvement in andcommitment to recreational tennis suggested that these two concepts arehighly correlated but distinct. Backman and Crompton (1991) found a sig-nificant relationship between involvement and loyalty among golfers and ten-nis players. Using data from participants of weight training and aerobicdance programs, Park (1996) found that involvement profiles and attitudinalloyalty profiles are significantly correlated and that involvement and attitu-dinal loyalty contribute independently to the prediction of different mea-sures of behavioral loyalty (i.e., duration, intensity, and frequency).

More recendy, Kim, Scott, and Crompton (1997) tested a model ex-plaining the interrelationships among involvement, commitment, and inten-tion to engage in birding trips and concluded that relationships did exist.They added, however, that the relationships were not fully understood, inpart because the effects of each independent variable was examined inde-pendently using regression analysis, not as part of an integrated model. Spe-cifically, Kim et al. concluded that "a more sophisticated model could readilybe developed" to show that "psychological commitment precedes both be-havioral involvement and commitment" (p. 338).

Definitions of key process variables. Involvement is defined as "an unob-servable state of motivation, arousal, or interest toward a recreational activityor associated product. It is evoked by a particular stimulus or situation andhas drive properties" (Havitz & Dimanche, 1997). In leisure research, "in-volvement has usually been treated as a multifaceted construct includingattraction, sign, centrality, and risk" (Havitz & Dimanche, 1999, p. 123). At-traction refers to the perceived importance or interest in an activity or aproduct, and pleasure or hedonic value derived from participation or use,whereas sign refers to the unspoken statements that purchase or participa-tion conveys about the person. Centrality encompasses both social contextssuch as friends and families centered around activities, and the central roleof the activities in an individual's life. Finally, risk consists of two primaryaspects: risk probability (perceived probability of making a poor choice) andrisk consequence (perceived importance of negative consequences in thecase of a poor choice).

Pritchard et al. (1999) defined psychological commitment as a tendencyto resist change. They argued that formative processes (i.e., informational,identification, and volitional) activate this tendency. Informational processesare represented, first, by informational complexity. This is the degree of

50 IWASAKI AND HAV1TZ

complexity of a person's cognitive structure (McQuiston, 1989). Informa-tional processes are also defined in terms of cognitive consistency, or thecongruence between beliefs or values and attitudes (Rosenburg, 1960). Fi-nally, informational processes are represented in terms of confidence or de-gree of certainty associated with attitudes and/or behaviors (Berger & Mitch-ell, 1989). Position involvement (the degree to which self-image is linked tobrand preference; Freedman, 1964) and volitional choice (the extent towhich a decision to perform an action is based on a person's free choice;Bagozzi, 1993) play a key role in the identification and volitional processes,respectively (Pritchard et al., 1999).

Consistent with Crosby and Taylor's (1983) definition of psychologicalcommitment, Pritchard et al. (1999) postulated that "resistance to change,as the principal evidence of commitment, will act as a mediator between theconstruct's antecedent processes and loyalty" (p. 337). Resistance to changerefers to individuals' unwillingness to change their preferences toward, im-portant associations with, and/or beliefs about a brand (e.g., a product, anagency).

Researchers generally agree that loyalty measures should combine bothbehavioral and attitudinal components (e.g., Backman & Crompton, 1991;Dick & Basu, 1994; Howard et al., 1988). Behavioral loyalty is comprised ofseveral components, including duration (long-term length of participation,patronage, or use), frequency (number of purchases, uses, or participationover a specified time-period; for example, a week, month, season, or year),intensity (hours per week or days per month devoted to purchase, use, orparticipation), sequence (purchase patterns within or between brands), pro-portion (the percentage of brand loyalty), and probability of brand use overtime (its intent being to predict future behavioral loyalty; Havitz 8c Howard,1995; Park, 1996). Attitudinal loyalty is reflected in the components of psy-chological commitment described above (Pritchard et al., 1999).2 Thus, Iwa-saki and Havitz's model has incorporated both attitudinal and behavioralaspects of loyalty.

Progressive processes leading to behavioral loyalty. "Leisure involvement re-fers to how we think about our leisure and recreation" (Havitz & Dimanche,1997, p. 246), whereas commitment represents "consumer attitudes of at-tachment to a brand" (Pritchard et al., 1999, p. 334). On the other hand,loyalty has been defined as "a composite blend of brand attitude and be-havior" (e.g., Day, 1969; Pritchard & Howard, 1997). Therefore, the pro-cesses conceptualized in Iwasaki and Havitz's model represent aspects of be-liefs, attitudes, and behaviors. Leisure involvement reflects people's beliefsabout their leisure participation/activity, whereas psychological commitmentand attitudinal loyalty reflect their attitude toward a brand of service. In

2Detailed information for a rationale for such consideration of attitudinal loyally can be foundin Iwasaki and Havitz (1998; pp. 271-272).


contrast, behavioral loyalty represents people's behavior in their leisure (e.g.,the use of a recreation service provider).

It has been widely shown that "beliefs play a central role in attitudetheory because they provide the groundwork upon which attitudes areconstructed. . . Thus, beliefs seen as being important and consistent lead tostrong attitudes" (Madrigal, 2001, p. 149). Consequently, attitudes consid-ered as being important play a key role when individuals process informa-tion, form intentions, and take actions (Boninger, Krosnick, & Berent, 1995;Fishbein 8c Ajzen, 1975). A belief-attitude-behavior linkage/hierarchy hasbeen both theoretically and empirically established in the past (e.g., Ajzen,1991, 2000; Albarracin, Johnson, Fishbein, & Muellerleile, 2001). It is im-portant to emphasize that the progressive processes identified in Iwasakiand Havitz's model are consistent with this belief-attitude-behavior linkage/hierarchy.

Personal and social factors that may influence involvement-loyalty relationships.As well as considering psychological commitment as a key mediating variablelinking involvement to behavioral loyalty, Iwasaki and Havitz's (1998) modelsuggests that personal and social factors may influence the relationship be-tween involvement and loyalty. For example, skill/competence and motiva-tion are assumed to be key personal antecedents of involvement and keypersonal moderators, whereas social support and social norms are consid-ered as key social antecedents of involvement and key social moderators. InSiegenthaler and Lam's (1992) study on recreational tennis, skill was foundto be significantly and positively related to ego-involvement (i.e., "the iden-tification of self with an activity," Siegenthaler & Lam, 1992, p. 304). Also,Backman and Crompton (1991) found that competence/skill of golfers andtennis players significantly contributed to differentiating loyalty categories.According to Kuentzel and McDonald's (1992) study on river use, speciali-zation dimensions (including commitment) were found to be significantlyand differently related to attitudinal indicators such as motives. Furthermore,Gammonley (2001) found evidence for a positive association between in-volvement in recreation and community activities and the enhancement ofsocial support, whereas Johnston and Carroll (2000) showed that involve-ment in sport was positively associated with the use of a support-seekingcoping strategy during rehabilitation processes among patients who had sus-tained injury restricting their normal functioning. In Dick and Basu's (1994)conceptual framework of client loyalty, social norms are identified as mod-erators of loyalty.

Other personal or social factors such as satisfaction and side bets/sunkcosts likely influence the relationship between involvement and loyalty, onceindividuals have developed their involvement in an activity. Allen, Machleit,and Schultz Kleine (1992) have shown the moderating effects of satisfactionand emotion on loyalty, while Dick and Basu (1994) conceptualized satisfac-tion as one of the key affective antecedents of loyalty. Also, Backman andCrompton (1991) found that side bets significantly discriminated loyalty cat-

52 IWASAKI AND HAVITZ

egories, and sunk costs and switching costs are identified as key conative/behavioral antecedents of loyalty in Dick and Basu's (1994) model.

Purpose and Research Questions

The purpose of the present study was to examine mediating effects ofpsychological commitment on the relationship between leisure involvementand behavioral loyalty to a recreation agency. Concomitantly, the moderatingeffects of selected personal and social factors on this belief-attitude-behaviorsequence were also examined. Testing these mediating effects may contributeto better understanding the processes by which clients become loyal to arecreation agency, while testing moderating effects helps further understandpersonal and social factors that influence these processes. This research isground-breaking in the sense that past research has not focused on the linkbetween all three factors—leisure involvement, commitment or loyalty (e.g.,Kim et al., 1997; Park, 1996).

Specific research questions are: (a) Does psychological commitment me-diate the effects of leisure involvement on behavioral loyalty? (b) Does resis-tance to change mediate the effects of commitment's formative factors onbehavioral loyalty? and (c) Do selected personal and social factors directlyinfluence leisure involvement, or moderate the involvement-behavioral loy-alty relationships? As suggested by past research, the personal and socialfactors examined in this study are: (a) skill/competence, (b) motivation, (c)social support, (d) social norms, (e) satisfaction, and (f) side bets/sunk costs.

There are, of course, leisure contexts wherein repeat visitation may notbe viewed as a desirable management outcome, for example when publicland resources are stressed or when overcrowding exacerbates user-groupconflicts (McLean, Havitz, & Adkins, 2002; Manning, 1999; Sem & Vogt,1997). Nevertheless, the development of participant loyalty has been recog-nized as an important organizational goal in many other recreation settings(Backman & Crompton, 1991; Bullaro & Edginton, 1986; Gahwiler & Havitz,1998; Howard et al., 1988). Loyal clients help recreation agencies meet prag-matic goals such as increased or stable revenues, generating a positive rep-utation, enhancing the capacity for networking, and several important socialand political goals (e.g., enhancing citizens' quality of life, developing com-munity, and increasing participation rates). In addition, loyalty may offerbenefits for participants/clients themselves (Anderson & Sullivan, 1993;Oliva, Oliver, & MacMillan, 1992). For example, loyal participants who ex-hibit strong interest in a leisure activity and are committed to a recreationagency, appear to experience personal goals (e.g., improved skills, health,quality of life, self-identity, self-expression, self-actualization) and social re-wards from that association (e.g., satisfying social relationships, a sense ofbelonging, social identity, group accomplishment; Csikszentmihalyi, 1990;Kaczynski 8c Havitz, 2001; Stebbins, 1998, 2000).

Thus, understanding the characteristics of loyal leisure participation of-fers an important challenge to the field. It is particularly helpful for recre-ation service managers to understand how and why clients develop their


loyalty to a recreation agency. The method discussion that follows explainsthe approach used to explore this important question.

Methods

Data Collection

Prior to the collection of data, the principal investigator contacted thedirectors of recreation service agencies and secured cooperation. The tworesearch assistants of the study systematically intercepted clients (every thirdperson) around reception areas at several locations of these agencies duringvarious times of days (mornings, afternoons, and evenings) on one weekdayand one weekend. The research assistants were rotated to perform their tasksfor subject recruitment for an entire week so that a systematic bias was re-duced. The locations were selected to represent different sections of the city(i.e., central, northeast, northwest, southeast, and southwest). Those clientsintercepted were briefly explained the purpose and nature of the study andasked to participate in the study. Those who agreed to participate broughthome a project package (which consisted of a cover letter describing thepurpose and importance of the study, instructions summarizing their tasks,a set of questionnaires including the measures to be described below, and aself-addressed and stamped envelope). They followed the instructions andreturned their completed questionnaires to the principal investigator usinga self-addressed and stamped envelope. Those participants who returnedtheir completed questionnaires received $10 in appreciation of their timeand contribution to the study. Payment was deemed appropriate given thefairly extensive (8-page long) questionnaire.

Measures

Leisure involvement. The participants' levels of involvement were mea-sured using a modified version of Laurent and Kapferer's (1985) ConsumerInvolvement Profile (CIP) scale, with respect to the participants' most favor-ite leisure activities. The scale consisted of three facets for enduring involve-ment: (a) attraction (6 items), (b) sign (3 items), and (c) centrality (3 items),and two facets for risk involvement: (a) risk probability and (b) risk conse-quence (3 items each). The addition of the centrality facet to enduring in-volvement was examined, consistent with recent leisure and involvement re-search (e.g., Mclntyre, 1989; Schuett, 1993). The measure used a five-pointLikert-type scale (1 = "strongly disagree" to 5 = "strongly agree"). The CIPscale has been widely used in leisure and involvement research and has beenshown to have good validity and reliability (see Havitz & Dimanche, 1997).In the present study, risk items were prefaced with an explanatory sentenceand followed-up with screening questions in order to improve the reliabilityand validity of those facets.

Psychological commitment. Participants' levels of psychological commit-ment were assessed by measures of the construct's formative factors and par-ticipants' resistance to change. These scales were derived from Pritchard et

5 4 IWASAKI AND HAVTTZ

al.'s (1999) Psychological Commitment Instrument (PCI), and were adaptedto the context of participants' primary recreation service providers. The PCIconsisted of 3 items for each of four facets: (a) informational complexity,(b) volitional choice, (c) position involvement, and (d) resistance to change.Pritchard et al. found satisfactory psychometric properties of the PCI. Theoriginal Iwasaki and Havitz (1998) model, based on Pritchard et al.'s (1992)original work, proposed six facets of psychological commitment. However, asa result of extensive testing for scale purification, validity, and reliability usingdata from travel service users, Pritchard et al. (1999) suggested droppingcognitive consistency and confidence, as the two facets failed to converge asdiscrete components and tended to be reflected in the informational pro-cesses of psychological commitment. As pointed out previously, Pritchard etal.'s analyses have suggested that resistance to change is a key precursor toloyalty that is distinct from other formative components of psychologicalcommitment.

Behavioral loyalty. Participant behavioral loyalty was captured by two ma-jor behavioral aspects of leisure participation at a public recreation facility:frequency of attendance and proportion of participation. Frequency was as-sessed by asking: "On average, how many days a week did you participate at

(primary recreation service agency)?" Two items were used formeasuring proportion of participation: "How many hours in a typical weekdo you spend at (primary recreation service agency)?" and "In-cluding the hours you spend at (primary recreation serviceagency), how many hours in total do you participate in recreational andsocial activities outside your home in a typical week?" Participants' responsesto the former were divided by their responses to the latter to calculate pro-portion of participation at their primary recreation service agencies. Al-though other forms of behavioral measures such as duration (length of par-ticipation, patronage, or use), sequence (purchase patterns within orbetween brands), and probability of brand use over time could be used forassessing behavioral loyalty, frequency and proportion of participation ap-pear to reflect two primary aspects of behavioral loyalty (Havitz & Howard,1995; Park, 1996).

Personal and social moderators. Skill/competence was assessed using a single-item self-rated measure of skill (Kuentzel & Heberlein, 1997; five-point scale;1 = none, 2 = novice, 3 = intermediate, 4 = high, and 5 = expert) withrespect to the participants' most favorite leisure activities. Their motivationtoward such activities was measured using Carroll and Alexandris' (1997)Strength of Motivation Scale (five-point scale; 1 = strongly disagree to 5 =strongly agree) consisting of five items (e.g., "I regret when I am unable toparticipate in "). The level of social support was assessed by thefour items for measuring emotional social support included in Iwasaki andMannell's (2000) Leisure Coping Scale (five-point scale; 1 = strongly disa-gree to 5 = strongly agree; e.g., "I feel emotionally supported by my leisurecompanions."). Whereas, social norms toward most favorite leisure activitieswere measured by Ajzen and Driver's (1992) two-item scale of subjective


norms (five-point scale; 1 = strongly disagree to 5 = strongly agree;e.g., "Most people who are important in my life think I should engage in

")A single-item five-point scale was used to measure the participants' sat-

isfaction with their most favorite leisure activities ("Overall, I would rate mylevel of satisfaction with as: 1 = very low to 5 = very high").Finally, side bets/sunk costs toward such activities were measured using six-itemfive-point scales (1 = strongly disagree to 5 = strongly agree) developedbased on Buchanan (1985) and Dick and Basu (1994). A sample item is that"I have invested a lot of money in (e-g-> costs, equipment, mem-bership, lessons)."

Analysis

First, descriptive analyses of the measures used were performed. Then,confirmatory factor analysis (CFA) using AMOS 4.0 (Arbuckle & Wothke,1999), with a maximum likelihood estimation method, was carried out totest goodness of fit of the measurement models for the constructs examined,and the factor structures and dimensionalities of these constructs. Conse-quently, structural equation modeling (SEM) using AMOS 4.0, with a maxi-mum likelihood estimation method, was conducted to examine the mediat-ing effects of psychological commitment. To test these mediating effects, wecompared goodness of fit of the Fully Mediated Model (F-M-M; see Figure2) with goodness of fit of the Direct Effects Models (D-E-M, also see Figure2). The two D-E-M models were not developed at random or simply pre-sented as "straw men" clearly inferior to the F-M-M. Rather, both representedmodels deemed viable within the leisure literature. The F-M-M assumes thatinvolvement indirectly influences behavioral loyalty via psychological com-mitment. In contrast, the Direct Effects Model I (D-E-M I) included thedirect path from involvement to behavioral loyalty in addition to the medi-ating paths depicted in the F-M-M. The direct path from involvement tobehavioral loyalty represents the dominant mode of thinking among leisureinvolvement researchers in the early 1990s (see Havitz & Dimanche, 1999for a summary of this literature). Likewise, the Direct Effects Model II (D-E-M II) included the direct path from commitment's formative factors to be-havioral loyalty, a relationship postulated by Pritchard (1992) in his originalresearch.

The comparison of the D-E-Ms with the F-M-M allowed us to test whetherpsychological commitment mediates the effects of leisure involvement onbehavioral loyalty (Research Question 1) and whether resistance to changemediates the effects of commitment's formative factors on behavioral loyalty(Research Question 2). These procedures are consistent with a current stan-dard practice of SEM that moves beyond simply testing a proposed modelby comparing the performance of the proposed model with competing orrival models (Arbuckle & Wothke, 1999; Hubbard & Mannell, 2001). Eachof the three models consisted of the same four constructs (i.e., involvement,

56 IWASAKI AND HAVTTZ

Fully Mediate Model (F-M-M)

Commitment sFormative Factors

Direct Effects Model I (D-E-M I)

Direct Effects Model II (D-E-MII)

Figure 2. Fully Mediated Model (F-M-M) and Direct Effects Models (D-E-Ms) Tested

commitment's formative factors, resistance to change, and behavioral loy-alty), and the differences among the models were attributed to the specifi-cation or non-specification of a certain path between constructs, as describedabove. Thus, these are considered nested models. Pedhazur and Schmelkin(1991) stated, "When one or more free parameters of a model are con-strained (e.g., constraining them to equal zero), the model thus obtained issaid to be nested in the one from which it was derived. A nested model,therefore, is a 'special case' . . . or a 'specialization' . . . of a more compre-hensive model" (p. 651). To perform significance tests for these nested mod-els, we followed a protocol described by Bollen (1989, p. 289-292). Specifi-cally, we used the likelihood ratio (LR) test. Bollen (1989) argued, "Inpractice, analysts calculate the LR statistic as the difference in the usual chi-


square estimators for the restricted and unrestricted models, with df equalto their difference in df. For this reason the LR test is often called a chi-square difference test" (p. 292). Furthermore, to examine the moderatingeffects of selected personal and social factors on involvement-loyalty rela-tionships (the second part of Research Question 3), we tested the best modelfound in the previous analyses in different conditions (e.g., high skill vs. lowskill groups), using medium splits. Finally, regression analyses were used totest the direct effects of personal and social factors on leisure involvement(the first part of Research Question 3).

Results

Participants

The sample consisted of 296 individuals who were users of fitness facil-ities or participants in fitness-related programs at one of two recreation ser-vice agencies in a large western Canadian city. Of 166 female and 128 maleparticipants,3 ages ranged from 18-25 (n = 110) to 26-35 (n = 64), 36-45(n = 64), 46-55 (n = 41), 56-65 (ra = 11), and 66-75 (n = 4). Most of theparticipants were Caucasian (n = 239). Other racial groups included Asian(ra = 33), African (ra = 7), Aboriginal (ra = 2), and South American (ra =2). The participants were predominantly single (ra = 145) or married (w =122), but some were divorced (ra = 12), separated (ra = 3), common law(n = 3), engaged (ra = 4), and widowed (ra = 1). The majority of the par-ticipants were either paid workers (full-time, n = 153; part-time, n = 47) orstudents (full-time, n = 64; part-time, n = 8), and other occupational cate-gories included retired (n = 6), household workers (ra = 5), unemployed(ra = 2), and semi-retired (ra = 1). Their annual incomes in Canadian dollarsranged from none (ra = 1) and below $10,000 (ra = 87) to $10,001-$30,000(n = 81), $30,001-$50,000 (ra = 83), $50,001-$70,000 (ra = 22), $70,001-

3,000 (ra = 7), and more than $90,001 (n = 5).

Descriptive Statistics

Table 1 presents correlation coefficients and descriptive statistics of themeasures used. As expected, all the measures representing key constructs/factors (i.e., enduring involvement, risk involvement, psychological commit-ment, and behavioral loyalty) were positively and significandy correlated witheach other (e.g., attraction, sign, and centrality for enduring involvement).Although risk consequence (as a facet of risk involvement) was positively andsignificantly correlated with sign (as a facet of enduring involvement), riskprobability (as a facet of risk involvement) had a negative and significantcorrelation with all the enduring involvement measures. This observation isnot surprising and is consistent with most leisure involvement research (Hav-

individuals did not indicate their gender.

TABLE 1Correlation Coefficients and Descriptive Statistics of Measures Used

00

10 11 12 13 14 15 16 17 18 19

1. Attraction (enduringinvolvement)

2. Sign (enduring involvement)3. Centrality (enduring

involvement)4. Risk probability (risk

involvement)5. Risk consequence (risk

involvement)6. Informational complexity

(commitment's formativefactors)

7. Volitional choice(commitment's formativefactors)

8. Position involvement(commitment's formativefactors)

9. Resistance to change item 110. Resistance to change item 211. Resistance to change item 312. Frequency (behavioral loyalty)13. Proportion (behavioral loyalty)14. Skill15. Motivation16. Social support17. Social norms18. Satisfaction19. Side bets/sunk costs

.43*

.31*

- .03

.16*

.56*

- .41* -.25* -.19*

.13* .08 .14*

.23* .20* .25* -.17* .04

.02 -.02 -.14* .001 .56*

.18* .32* .20* -.14* .13* .49* .47*

.17*

.11

.15*

.07-.07

.12*

.50*

.19*

.28*

.36*

.25*

.16*

.05

.12*

.08

.02

.19*

.36*

.36*

.18*

.29*

.31*

.12-.01-.02

.18*

.03

.33*

.28*

.47*

.16*

.23*

.45*

-.13*-.12-.14*-.10-.03-.05-.35*-.09-.24*-.25*-.09

.09

.10

.15*-.02-.03

.04

.11

.07

.02

.01

.18*

.61*

.45*

.45*

.31*

.11

.31*

.17*

.19*

.19*

.19*

.29*

.47*

.44*

.46*

.17*

.12

.09

.08

.05

.15*

.04-.02

.53*

.47*

.50*

.18*

.07

.09

.10

.25*

.14*

.14*

.16*

.49*

.54*

.21*

.08

.09

.23*

.10

.08

.12*

.15*

.72*

.22*

.15*

.02

.13*-.02-.02

.12*

.06

.21*

.12*

.06

.22*

.02

.09

.17*

.12*

.26*

.13*

.10

.08

.04

.04

.16*

.002-.08-.09-.08-.02

.08

.13*

.21*

.18*

.27*

.30*

.16*

.22*

.37*

.24*

.20*

.13* .20*

.24* .13* .17*

Means1

Standard deviations4.15

.563.12.91

2.821.03

2.02

.632.42 3.65.97 1.02

3.89 2.60 3.331.06 1.10 1.17

3.16 3.30 3.381.37 1.42 2.22

.50 3.49 4.09 3.08 4.21 4.32 3.31

.28 .76 .67 .95 .82 .68 .73

Notes: *p < .05'Means for all measures, excluding proportion of purchase, are generated on a l-to-5 scale.


itz 8c Dimanche, 1997). Risk probability is largely a confidence measure, andthe more involved a participant is (on other facets), it makes conceptualsense that the lower are her or his chances of making a poor choice.

Skill, motivation, social support, social norms, satisfaction, and side betsall had positive and significant associations with each of the three indicatorsof enduring involvement and with informational complexity. Social support,social norms, satisfaction, and side bets were positively and significantly as-sociated with position involvement. By contrast, motivation, social norms,and satisfaction were negatively and significantly correlated with risk proba-bility. Motivation and satisfaction showed a positive and significant correla-tion with all the indicators of resistance to change, and side bets with two ofthe three indicators of resistance to change. Skill and side bets were positivelyand significantly associated with the frequency measure of behavioral loyalty.

Measurement Models: Confirmatory Factor Analysis

Measurement models for all the constructs/factors to be used for thesubsequent structural equation modeling (SEM) were created, and goodnessof fit of these models was tested using confirmatory factor analysis (CFA).Because risk probability was negatively and significantly correlated with allthe enduring involvement measures, separate measurement models were cre-ated for risk involvement and enduring involvement. The measured varia-bles/indicators depicted with squared boxes in Figure 3 represent partici-pants' mean scores of the items for enduring involvement, risk involvement,and psychological commitment, and their raw scores of the three question-naire items for resistance to change. Behavioral loyalty was comprised offrequency and proportion indicators with respect to the use of primary rec-reation agencies.

The results of CFA using AMOS 4.0 are presented in Figure 3 with theassumption that each of the five constructs/factors is correlated with eachother. As for the selection of overall fit indices for CFA and SEM with max-imum likelihood (ML) estimation methods, Hu and Bentler's (1998) evalu-ation of the sensitivity of various fit indices to model misspecification hasrecommended the use of standardized root-mean-square residual (SRMR),supplemented by Tucker-Lewis Index (TLI), Bollen's fit index (BL89), rela-tive noncentrality index (RNI), comparative fit index (CFI), gamma hat,McDonald's centrality index (Me), or root-mean-square error of approxi-mation (RMSEA). However, no definite recommendation has been provided.Bollen (1989) suggested that "the safest recommendation is to always reportthe chi-square estimate along with several of the other fit indices (e.g., re-siduals; normed fit index, NFI; incremental fit index, IFI; relative fit index,RFI; and Tucker-Lewis Index, TLI)" (p. 281). In contrast, Hu and Bender(1998) pointed out that "as noted by many researchers, . . . the standardchi-square test may not be a good enough guide to model adequacy" (p.425). The selection and interpretation of overall fit indices reported in Table2 were based on the above arguments. Generally, the measurement models


51

62

83

54

85

86

87

88

89

810

811

812

513

.63

.46

.30

.11

.58

.46

.49

.65

.69

.51

.13

Attraction

Sign

Centrality

Risk Probability

Risk Consequence

Informational Complexity

Volitional Choice

Position Involvement

RC Item 1

RC Item 2

RC Item 3

Frequency

Proportion

^ 5 4

^ 7 6

.68 f

• ^69

^- --xEnduring

Involvement

^ ^

^ ^Risk

Involvement

^ ^

^ "̂ NPsychologicalCommitment

^ ^

Resistance toChange

^ . ^

^ ^ XBehavioral

Loyalty

" ^ ^

-.46 \ \ >

\+y .35 \

\

\-.26 y \

\

.14

.83 1 X /

N Jy i

.19

-.21 /

J*^^ .44| / /

.38 / / /

Note: Correlation coefficients estimated between factors are indicated with double-arrows. The numbersindicated on the right of each of the 13 measured variables/indicators (depicted with squared boxes) arestandardized factor loadings, whereas the numbers indicated on the left of these variables/indicators are squaredmultiple correlations that reflect the amount of each indicator's variance explained by its respective factor. 8 =delta (measurement errors).Model fit indices: •£ = 215.17 (df = 56, p < .01); Normed Fit Index (NFI) = .98; Relative Fit Index (RFI) =.96; Increment Fit Index (IFI) = .98; Tucker-Lewis Index (TLI) = .97; Comparative Fit Index (CFI) =.98; Standardized Root Mean Residual (SRMR) = .09

Figure 3. Measurement Models Tested Using Confirmatory Factor Analysis (CFA)

fit to the data fairly well. As for component fit indices, all of the indicatorsloaded significantly on their designated factors (p < .01).

The indicators used in the present study showed acceptable to moder-ately high internal consistency in measuring key constructs. Composite reli-ability estimates were .75, .63, .59, and .56 for enduring involvement, com-mitment's formative factors, resistance to change, and behavioral loyalty,respectively. Correlations between the five factors estimated are shown inFigure 3. A high correlation between commitment's formative factors and


TABLE 2Analysis of Competing Structural Models

Direct effects ((3s):Enduring Involvement —>

Commitment's Formative ProcessesCommitment's Formative Processes —>

Resistance to ChangeResistance to Change —* Behavioral

LoyaltyEnduring Involvement —* Behavioral

LoyaltyCommitment's Formative Processes —•

Behavioral Loyalty

Squared Multiple Correlations (R2):Commitment's Formative ProcessesResistance to ChangeBehavioral Loyalty

Model fit indices:Chi-Sqaure (df)Normed Fit IndexRelative Fit IndexIncrement Fit IndexTucker-Lewis IndexComparative Fit IndexStandardized Root Mean Residual

F-M-M

.28*

.83*

.44*

.08*

.69*

.19*

159.95* (41).98.97.99.98.99.09

D-E-M I

.28*

.83*

.40*

. l l n s

.08*

.69*

.19*

158.63* (40).98.97.99.98.99.10

D-E-M II

.28*

.82*

.29*

14ns.

.08*

.67*

.20*

156.75* (40).98.97.99.98.99.10

Notes: *p < .01; n.s. = not statistically significant at .05 level

resistance to change raised some concern with the discriminant validity be-tween these two constructs. As recommended by Burnkrant and Page (1982),we compared the existing measurement model (x2 = 215.17, df = 56, p <.01) in which all the five factors were allowed to correlate, with one in whichcommitment's formative factors and resistance to change were hypothesizedto have a unity correlation that depicted as unidimensional (x2 = 238.99, df= 57, p < .01). A chi-square difference test between the two models sup-ported the existing model (Figure 3) and suggested that the two constructsshould be considered as discrete factors, showing the discriminant validitybetween these constructs (Ax2 = 23.82, df = 1, p < .01). This finding isconsistent with data reported by Pritchard et al. (1999).

Structure of Involvement-Loyalty Relationships: Structural Equation Modeling

Next, we performed structural equation modeling (SEM) analyses to testthe Fully Mediated Model (F-M-M; Figure 2) in comparison to the DirectEffects Models I and II (D-E-M; Figure 2). As noted earlier, because of the


negative and significant association between enduring involvement and riskprobability (as a facet of risk involvement), testing of the F-M-M began withhaving both enduring involvement and risk involvement as antecedents ofpsychological commitment. However, we found that the effects of risk in-volvement on both psychological commitment and behavioral loyalty werenot statistically significant. The problem of risk involvement with respect toits conceptualization and measurement has been recognized in the past. Ac-cording to Havitz and Dimanche's (1997) review of evidence from 50 leisureinvolvement data sets, one of the reoccurring results has been a poor per-formance of risk involvement items. They have identified challenges in ap-propriately conceptualizing and measuring risk involvement and have sug-gested that refined measures of risk involvement are needed. Thus, riskinvolvement was dropped in the further analyses. Table 2 presents the resultsof overall fit indices, beta values, and squared multiple correlations (SMC;indicating the amount of variance explained) for the F-M-M without riskinvolvement. Overall, the results suggested a good fit of the F-M-M to thedata. All of the mediating paths leading to behavioral loyalty were found tobe statistically significant. The amount of variance explained by enduringinvolvement in predicting commitment's formative factors was 8%, whereas69% and 19% of the variance in predicting resistance to change and behav-ioral loyalty were explained by commitment's formative factors and resistanceto change, respectively.

Competing or rival models that assumed direct links from enduring in-volvement to behavioral loyalty (D-E-M I) and from commitment's formativeprocesses to behavioral loyalty (D-E-M II) were also tested. Table 2 reportsoverall and component fit indices of both models. A LR or chi-square dif-ference test suggested that the addition of the direct path from enduringinvolvement to behavioral loyalty in the D-E-M I did not improve goodnessof fit, in comparison to the more restrictive F-M-M (Ax2 = 1.32, df= 1, p >.05). The path from enduring involvement to behavioral loyalty was foundto be not statistically significant (P = .11, p > .05; Table 2). Thus, the datalead us to conclude that the F-M-M is a better description of the relationshipsbetween enduring involvement and behavioral loyalty than the D-E-M I. Thatis, enduring involvement indirectly influenced behavioral loyally with psy-chological commitment mediating the effect of enduring involvement onbehavioral loyalty.

Similarly, the addition of the direct path from commitment's formativefactors to behavioral loyalty in the D-E-M II did not improve goodness of fit,in comparison to the more restrictive F-M-M (A\2 = 3.20, df - 1, p > .05).We found that the path from commitment's formative factors to behavioralloyalty was not statistically significant ((3 = .14, p > .05; Table 2). Therefore,the data suggest that the F-M-M is a better model describing the relationshipsbetween enduring involvement and behavioral loyalty than the D-E-M II.Specifically, commitment's formative factors indirectly influenced behavioralloyalty, and resistance to change mediated the effect of commitment's for-mative factors on behavioral loyalty.


Testing for mediation. In addition to the above LR or chi-square differ-ence tests, our review of the conditions for mediation (Baron & Kenny, 1986)suggested that the mediating effects of psychological commitment were in-deed present. As for the mediating effects on the relationship between en-during involvement and behavioral loyalty, first, the F-M-M showed that en-during involvement had a significant effect on commitment's formativefactors. Second, the F-M-M also noted that commitment's formative factorshad a significant effect on resistance to change, as well as a significant effectof resistance to change on behavioral loyalty. Third, when commitment'sformative factors and resistance to change were constrained (i.e., not linkedto behavioral loyalty), enduring involvement had a significant effect on be-havioral loyalty (beta = .15, p < .05). Fourth, the D-E-M I showed that thepreviously significant effect of enduring involvement on behavioral loyaltybecame non-significant or was significantly reduced (p > .05) when the me-diating paths from enduring involvement, through psychological commit-ment, to behavioral loyalty were opened.

Similarly, the potential mediating effects of resistance to change on therelationship between commitment's formative factors and behavioral loyaltywere examined. First, the F-M-M indicated that commitment's formative fac-tors had a significant effect on resistance to change. Second, the F-M-M alsoshowed that resistance to change significantly influenced behavioral loyalty.Third, when resistance to change was constrained (i.e., not linked to behav-ioral loyalty), commitment's formative factors had a significant effect on be-havioral loyalty (beta = .44, p < .001). Fourth, the D-E-M II showed that thepreviously significant effect of commitment's formative factors on behavioralloyalty became non-significant (p > .05) when the mediating paths fromcommitment's formative factors, through resistance to change, to behavioralloyalty were opened.

Direct and Moderating Effects of Personal and Social Factors

Next, to examine the direct effects of personal and social factors onenduring involvement, four selected personal and social antecedent factors(i.e., skill, motivation, social support, and social norms) were simultaneouslyentered into a regression model in predicting enduring involvement. Eachof these antecedent factors was found to significantly predict enduring in-volvement (R2 = .40). Greater levels of skill ((3 = .13, p < .05), motivation(P = .41, p < .05), social support ((3 = .32, p < .05), and social norms ((3= .11, p < .05) were significantly related to higher levels of enduring in-volvement.

Finally, to examine the moderating effects of personal and social factorson the relationships between enduring involvement and behavioral loyalty,the F-M-M was tested separately for different conditions (e.g., high skillgroup versus low skill group), using the medium splits. The results of theseanalyses are summarized in Table 3. Of the six potential moderators tested,we found skill, motivation, social support, and side bets all moderated the


TABLE 3Analysis of Moderating Effects of Selected Personal and Social Factors

Enduring Involvement Commitment's Resistance to Change—» Commitment's Formative Processes —» —» Behavioral

Formative Processes Resistance to Change Loyalty

Skill:Low (n = 148)High (n = 148)

Motivation:Low (n = 149)High (n = 147)

Social support:Low (n = 158)High (n = 138)

Social norms:Low (n = 156)High (n = 140)

SatisfactionLow (n = 157)High (n = 129)

Side bets/sunk costsLow (n = 145)High (n = 151)

.16ns-

.31**

1 7n. s .

.23*

- . 0 1 " s

.46**

.20*

.28*

.26*

.21*

.12ns-

.42**

.92**

.71**

.79**

.91**

.87**

.81**

.87**

.85**

.70**

.91**

.93**

.72**

.38**

.46**

.37**

.47**

.41**

.49**

.44**

.46**

.36**

.52**

.46**

.37**

Notes: Structural equation modeling was used to test the F-M-M separately for different condi-tions (e.g., high skill group versus low skill group), using the medium splits. **p < .01; *p <.05; n.s. = not statistically significant at .05 level.

effects of enduring involvement on commitment's formative factors. Specif-ically, in groups with low levels of these moderators the effects of enduringinvolvement on commitment's formative factors were found to be non-significant, whereas in groups with high levels of the moderators these effectswere found to be significant. Higher levels of skill, motivation, social support,and side bets significantly accounted for a stronger relationship betweenenduring involvement and commitment's formative factors than did lowerlevels of these moderators. No other moderating effects were found to bestatistically significant.

Discussion

In this study we first examined mediating effects of psychological com-mitment on the relationship between leisure involvement and behavioralloyalty to a recreation agency, as hypothesized in Iwasaki and Havitz's (1998)revised model. Our study represents the first attempt to test comprehensivelythe relationships among three of the key variables having important impli-cations for leisure studies and services, namely, leisure involvement, psycho-


logical commitment, and loyalty. The findings of our analyses using confirm-atory factor analysis and structural equation modeling to test these mediatingeffects leading to behavioral loyalty contribute to better understanding theprocesses by which clients become loyal to a recreation agency. Furthermore,the results better explain ways in which different factors influence the rela-tionship between leisure involvement and behavioral loyalty. Below, the find-ings are summarized with respect to the three research questions examined.

The comparison of the Fully Mediated Model (F-M-M) with its rival mod-els, Direct Effects Models (D-E-M I & II), suggest that psychological com-mitment plays a mediating role in the relationship between enduring in-volvement and behavioral loyalty. Enduring leisure involvement does notappear to directly influence behavioral loyalty. Instead, enduring leisure in-volvement seems to indirectly influence behavioral loyalty via psychologicalcommitment. Also, it is important to distinguish two components of psycho-logical commitment, namely, commitment's formative factors and resistanceto change, since commitment's formative factors tend to indirectly influencebehavioral loyalty via resistance to change. The data suggest that currentconceptualizations involving mediation effects are more reflective of realitythan were previous models built primarily around direct effects.

These data suggest that not all highly involved leisure activity partici-pants become loyal to a recreation agency, although higher levels of endur-ing involvement seem to be an important precursor to behavioral loyalty ofthis type. Higher levels of psychological commitment, in which resistance tochange is a pivotal element, appear essential for the development of clientloyalty to a recreation agency. The development of client loyalty appears tobe best explained as a progressive process. That is, the formation of highinvolvement in a leisure activity seems to be a key precondition for becominga committed user of a leisure service agency and supporter of that agency.Loyalty, thus, seems to occur when people develop a resistance to change inpreferences, beliefs, and associations within the agency.

These findings are consistent with past research on leisure involvement,commitment, and/or loyalty. For example, James (2001) used Piaget's (1970)theory of cognitive development and Iwasaki and Havitz's (1998) model asthe conceptual framework to examine when and how children begin to dem-onstrate their loyalty to sports teams. Based on his interviews with childrenaged 5 to 6 (n = 25) and 8 to 9 (n = 25), James found that "half of thechildren interviewed . . . did have a favorite sports team, suggesting that theyhad an interest in or attraction to a specific team" (p. 256), and that thesechildren were capable of showing a psychological commitment to a favoriteteam. James suggested that "the current study does provide support for por-tions of the conceptual framework proposed by Iwasaki and Havitz (1998),particularly the relationship between involvement (attraction/preference),psychological commitment, and resistance to change" (p. 259).

According to Park's (1996) analyses of data obtained from 208 partici-pants of an adult fitness program, highly involved individuals tend to "con-tinue participation due to emotional attachment to and identification with


the program" (p. 246). Using data collected from 517 visitors to a birdingfestival, Kim et al. (1997) investigated relationships between involvement andcommitment. Their correlational analyses suggested that the facets of en-during involvement (what they termed "social psychological involvement"),namely, importance/pleasure and sign were significantly related to commit-ment, denned as "personal and behavioral investments that bind individualsto consistent patterns of leisure behavior" (p. 336). Kim et al. also foundthat "highly involved and committed birders tend to go birding often, traveland spend money on birding, are skilled at identifying birds, read aboutbirding, belong to birding organizations, and own equipment that facilitatesthe identification of birds" (p. 337).

More recently, Pritchard et al. (1999) examined the role of commitmentin the development of loyalty, using 421 airline and hotel patrons. The resultsof their path analyses suggested that "the tendency to resist changing pref-erence" is "a key precursor to loyalty, largely explained by a patron's willing-ness to identify with a brand" (p. 333). Similar to the findings of the presentstudy, the formative factors of psychological commitment (i.e., volitionalchoice, position involvement, and informational complexity) did not have asignificant direct effect on loyalty. Instead, resistance to change fully medi-ated the effects of commitment's formative factors on loyalty. Pritchard et al.emphasized that "our current study's rejection of a direct effects model sup-ports the notion that some formative processes are only indirectly related toloyalty. This suggests that a wider review of other loyalty antecedents couldbe undertaken and considered in the light of commitment's mediating ef-fect" (p. 345). These statements are consistent with the present findings withrespect to the non-significant direct paths from enduring involvement andcommitment's formative factors to behavioral loyalty. Most of the antecedentsof loyalty appear to have indirect effects on loyalty being potentially mediatedby factors such as individuals' resistance to change their preference of andattitude toward a brand.

The present study also examined the direct and moderating effects ofselected personal and social factors on the relationship between leisure in-volvement and behavioral loyalty. First, we found that skill, motivation, socialsupport, and social norms were significant predictors of enduring involve-ment. These findings are consistent with past research. For example, Siegen-thaler and Lam (1992) found that skills of recreational tennis players weresignificantly related to ego-involvement, while Ray's (1997) study providedempirical support for a conceptual framework depicting the relationshipsbetween enduring involvement in jogging and its motivating antecedents andbehavioral consequences. In her study to evaluate the effectiveness of inter-ventions focusing on peer education and advocacy through recreation andleadership, Gammonley (2001) showed that involvement in recreation andcommunity activities was positively associated with the enhancement of socialsupport. Similarly, Johnston and Carroll (2000) found evidence for a positiveassociation between involvement in sport and the use of a support-seekingcoping strategy during rehabilitation processes among patients who had sus-


tained injury restricting their normal functioning. Discussing social compar-ison theory in the context of aerobic exercise classes, Frederick, Havitz, andShaw (1994) suggested that individuals' involvement profiles are related to"the type and extent of social comparison made in recreational settings" (p.167) in which social norms appear to play a key role.

Furthermore, the present study provided some evidence supporting themoderating effects of personal and social factors. Of a number of moderat-ing effects examined, we found that skill, motivation, social support, and sidebets moderated the effects of enduring involvement on commitment's for-mative factors. Specifically, the data suggest that the positive relationshipbetween enduring involvement and commitment's formative factors isstronger for those individuals with higher levels of skill, motivation, socialsupport, and/or side bets associated with their leisure activities than for thosewith lower levels of these aspects of leisure activities. Once people becomehighly involved in leisure activities, greater levels of skill, motivation, socialsupport, and/or side bets appear to be key factors for developing individuals'psychological commitment toward their primary recreation service providers.Although examinations of moderator effects have rarely been conducted inleisure involvement and commitment research, these findings make sense incomparison to past research. For example, in their river-use study, Kuentzeland McDonald (1992) found that attitudinal indicators such as motives weresignificandy and differentiy related to specialization dimensions includingcommitment. Also, it has been suggested that individuals' investments or sidebets influence their commitment to brands such as an activity or agency (e.g.,Allen & Meyer, 1990; Buchanan, 1985; Meyer, Allen, & Smith, 1993). Forexample, Siegenthaler and Lam (1992) found that financial side bets towardtennis (i.e., tennis expenditures) were significantly related to both ego-involvement and commitment in tennis.

Overall, the findings from this study suggest that the relationship be-tween involvement and loyalty is very complex, and these findings help un-cover some of the key mechanisms/processes by which clients become loyalto a recreation agency. An understanding of these mechanisms/processeshas important implications for managers and professionals of recreation ser-vices. For example, marketing strategies may be developed to aim at strength-ening client loyalty by maximizing key antecedents of behavioral loyalty (i.e.,enduring involvement and psychological commitment), as well as personaland social factors/moderators (e.g., skill, motivation, social support, socialnorms, and side bets/sunk costs) conceptualized in Iwasaki and Havitz'smodel. However, recommending specific strategies for developing client loy-alty is premature at this stage because we did not attempt to segment par-ticipants and evaluate the effectiveness of different strategies in this study.Nevertheless, it appears intuitive to suggest, for example, that market seg-mentation using profiles of involvement, psychological commitment, and be-havioral loyalty to perform market analysis would be useful (Park, 1996).Based on such analysis, marketing strategies targeting and attracting a par-ticular segment of the market could be developed. For instance, Warrington


and Shim (2000) argued that, "when high-involvement consumers are satis-fied, they are expected to develop brand and store loyalties, and in doingso, represent important market segments. . . Marketing strategies designedto attract and retain these consumers can potentially lead to higher sales anda more satisfied core of committed customers" (p. 762). Kyle, Kerstetter, andGuadagnolo's (2002) research regarding 1 OK road race participants supportsthese general conclusions.

It should be noted that with the use of cross-sectional data we can notprovide definitive support for the existence of the causal links suggested bythe Fully Mediated Model (F-M-M) even though we found stronger supportfor the F-M-M in comparison to rival Direct Effects Models. However, the apriori specification of the model based on a sound theoretical rationale plusthe use of structural equation modeling (SEM) that allows a statistical testof the fit of the model to the data, provide clues about causal processes thanmore exploratory approaches (Bollen, 1989). Obviously, the use of a longi-tudinal design is needed in future research. Indeed, the present study wasdeveloped as the first stage of a longer-term project. Also, generalizability ofthe model must be examined using various population groups. Furthermore,as suggested by Pritchard et al. (1999), it is important to examine otherantecedent factors of loyalty since a large amount of variance in predictingbehavioral loyalty was left unexplained in the F-M-M.

Another important task of leisure researchers is to identify reasons whycertain individuals do not become involved participants and loyal clients.There is an important need for recreation service agencies to develop struc-tures (e.g., policies, human resources) and provide opportunities (e.g., pro-grams, services) for eliminating or reducing constraints associated with lei-sure participation and the use of recreation services. For example, singleworking parents with small children experience constraints unique to theirlife circumstances such as those associated with finance, family, accessibility,and time, which seem likely to inhibit both the development of enduringinvolvement with leisure activities and relationships and psychological com-mitment to leisure service providers. Public sector agencies, however, havepolitical and moral obligations to provide recreation services that accom-modate the needs of single working parents to the same extent that theystrive to serve perhaps more responsive loyal clients. Our conceptual modelof involvement-loyalty relationships identifies antecedents of involvementand moderating variables (including constraints) that help further explainthe complex relationship between involvement and behavioral loyalty. Test-ing of these antecedent factors and moderating effects has important theo-retical and practical implications, thus, needs to be carried out in futureresearch.

References

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human DecisionProcesses, 50, 179-211.


Ajzen, I. (2000). Nature and operation of attitudes. Annual Review of Psychology, 52, 27-58.Ajzen, I., & Driver, B. L. (1992). Application of the theory of planned behavior to leisure choice.

Journal of Leisure Research, 24, 207-224.Albarracin, D., Johnson, B. T., Fishbein, M., & Muellerleile, P. A. (2001). Theories of reasoned

action and planned behavior as models of condom use: A meta-analysis. Psychological Bulletin,127, 142-161.

Allen, C. T., Machleit, K. A., & Schultz Kleine, S. (1992). A comparison of attitudes and emotionsas predictors of behavior at diverse levels of behavioral experience. Journal of ConsumerResearch, /#(March), 493-504.

Allen, N. J., & Meyer, J. P. (1990). The measurement and antecedents of affective, continuance,and normative commitment to the organization. Journal of Occupational Psychology, 63, 1-8.

Anderson, E. W., & Sullivan, M. W. (1993). The antecedents and consequences of customersatisfaction for firms. Marketing Science, 12, 125-143.

Arbuckle, J. L., & Wothke, W. (1999). AMOS 4.0 user's guide. Chicago, IL: Smallwaters Corpora-tion.

Backman, S. J. (1991). Exploring the relationship between perceived constraints and loyalty.Journal of Leisure Research, 23, 332-344.

Backman, S. J., & Crompton, J. L. (1991). Differentiating between high, spurious, latent, andlow loyalty participants in two leisure services. Journal of Park and Recreation Administration,9(2), 1-17.

Backman, S. J., & Veldkamp, C. (1995). Examination of the relationship between service qualityand user loyalty. Journal of Park and Recreation Administration, 13(2), 29-212.

Bagozzi, R. P. (1993). On the neglect of volition in consumer research: A critique and proposal.Psychology and Marketing, i0(May/June), 215-237.

Barber, N., & Havitz, M. E. (2001). Canadian participation rates in ten sport and fitness activities.Journal of Sport Management, 15, 51-76.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in socialpsychological research: Conceptual, strategic, and statistical considerations. Journal of Person-ality and Social Psychology, 51, 1173-1182.

Beatty, S. E., Kahle, L. R., Be Homer, P. (1988). The involvement-commitment model: Theoryand implications. Journal of Business Research, 16, 149-167.

Berger, I. E., & Mitchell, A. A. (1989). The effects of advertising on attitude-accessibility, attitude-confidence, and attitude-behavior relationship. Journal of Consumer Research, 16, 269-279.

Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.Boninger, D. S., Krosnick, J. A., & Berent, M. K. (1995). The causes of attitude importance: Self-

interest, social identification, and values. Journal of Personality and Social Psychology, 68, 61-80.

Buchanan, T. (1985). Commitment and leisure behavior: A theoretical perspective. Leisure Sci-ences, 7, 401-420.

Bullaro, J., & Edginton, C. R. (1986). Commercial leisure services: Managing for profit, service, andpersonal satisfaction. New York: MacMillan.

Burnkrant, R. E., & Page Jr., T. J. (1982). An examination of the convergent, discriminant, andpredictive validity of Fishbein's behavioral intention model. Journal of Marketing Research,iP(November), 550-561.

Carroll, B., & Alexandris, K. (1997). Perception of constraints in strength of motivation: Theirrelationship to recreational sport participation in Greece. Journal of Leisure Research, 29, 279-299.

Crosby, L. A., & Taylor, J. R. (1983). Psychological commitment and its effects on postdecisionevaluation and preference stability among voters. Journal of Consumer Research, 9, 413-431.

Csikszentmihalyi, M. (1990). Flow: The psychology of optimal experience. New York: Harper Perennial.


Day, G. S. (1969). A two dimensional concept of brand loyalty. Journal of Advertising Research, 35,43-56.

Dick, A. S., & Basu, K. (1994). Customer loyalty: Toward an integrated conceptual framework.Journal of the Academy of Marketing Science, 22, 99-113.

Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introduction to theory andresearch. Reading, MA: Addison-Wesley.

Frederick, C. J., Havitz, M. E., & Shaw, S. M. (1994). Social comparison in aerobic exerciseclasses: Propositions for analyzing motives and participation. Leisure Sciences, 16, 161-176.

Freedman, J. L. (1964). Involvement, discrepancy and change. Journal of Abnormal and SocialPsychology, 69, 290-295.

Gahwiler, P., & Havitz, M. E. (1998). Toward a relational understanding of leisure social worlds,involvement, psychological commitment, and behavioral loyalty. Leisure Sciences, 20, 1-23.

Gammonley, D. (2001). Peer education and advocacy through recreation and leadership. Psy-chiatric Rehabilitation Journal, 25, 170-178.

Ganesh, J., Arnold, M. J., 8c Reynolds, K. E. (2000). Understanding the customer base of serviceproviders: An examination of the differences between switchers and stayers. Journal of Mar-keting, 64Qu\y), 65-87.

Havitz, M. E., & Dimanche, F. (1997). Leisure involvement revisited: Conceptual conundrumsand measurement advances. Journal of Leisure Research, 29, 245-278.

Havitz, M. E., & Dimanche, F. (1999). Leisure involvement revisited: Drive properties and par-adoxes. Journal of Leisure Research, 31, 122-149.

Havitz, M. E., & Howard, D. R. (1995). How enduring is enduring involvement? A seasonalexamination of three recreational activities. Journal of Consumer Psychology, 4, 255-276.

Howard, D. R., Edginton, C. R., & Selin, S. W. (1988). Determinants of program loyalty. Journalof Park and Recreation Administration, 6, 41-51.

Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity tounderparameterized model misspecification. Psychological Methods, 3, 424-453.

Hubbard, J., & Mannell, R. C. (2001). Testing competing models of the leisure constraint ne-gotiation process in a corporate employee recreation setting. Leisure Sciences, 23, 145-163.

Iwasaki, Y, & Havitz, M. E. (1998). A path analytic model of the relationships between involve-ment, psychological commitment, and loyalty. Journal of Leisure Research, 30, 256-280.

Iwasaki, Y, & Mannell, R. C. (2000). Hierarchical dimensions of leisure stress coping. LeisureSciences, 22, 163-181.

Jacoby, J., & Chestnut, R. W. (1978). Brand loyalty measurement and management. New York: Wiley.James, J. D. (2001). The role of cognitive development and socialization in the initial develop-

ment of team loyalty. Leisure Sciences, 23, 233-261.Johnston, L. H., & Carroll, D. (2000). Coping, social support, and injury: Changes over time

and the effects of level of sports involvement. Journal of Sport Rehabilitation, 9, 290-303.Kaczynski, A. T, & Havitz, M. E. (2001). Relational benefits in recreation services: Examining

differences between operating sectors. Journal of Park and Recreation Admninistration, 19(2),20-42.

Kelley, S. W., & Davis, M. A. (1994). Antecedents to customer expectations for service recovery.Journal of the Academy of Marketing Science, 22(winter), 52-61.

Kim, S. S., Scott, D., & Crompton,J. L. (1997). An exploration of the relationships among socialpsychological involvement, behavioral involvement, commitment, and future intentions inthe context of birdwatching. Journal of Leisure Research, 29, 320-341.

Kuentzel, W. F., & Heberlein, T. A. (1997). Social status, self-development, and the process ofsailing specialization. Journal of Leisure Research, 29, 300-319.

Kuentzel, W. F., & McDonald, G. D. (1992). Differential effects of past experience, commitment,and lifestyle dimensions on river use specialization. Journal of Leisure Research, 24, 269-287.


Kyle, G. X, Kerstetter, D. L., & Guadagnolo, F. B. (2002). Market segmentation using participantinvolvement profiles. Journal of Park and Recreation Administration, 20(1), 1-21.

Laurent, G., & Kapferer, J. N. (1985). Measuring consumer involvement profiles. Journal of Mar-keting Research, 22, 41-53.

Lee, B. A., & Zeiss, C. A. (1980). Behavioral commitment to the role of sport consumer: Anexploratory analysis. Sociology and Social Research, 64, 405-415.

Madrigal, R. (2001). Social identity effects in a belief-attitude-intensions hierarchy: Implicationsfor corporate sponsorship. Psychology and Marketing, 18, 145-165.

Manning, R. E. (1999). Crowding and carrying capacity in outdoor recreation: From normativestandards to standards of quality. In E. L. Jackson & T. L. Burton (Eds.) Leisure Studies:Prospects for the Twenty-first Century. State College, PA: Venture.

Mclntyre, N. (1989). The personal meaning of participation: Enduring involvement. Journal ofLeisure Research, 21, 167-179.

McLean, D. J., Havitz, M. E., & Adkins, K. D. (2002). Demarketing leisure services: The case ofmunicipal golf courses. Journal of Park and Recreation Administration, 20(2), 90-110.

McQuiston, D. H. (1989). Novelty, complexity, and importance as causal determinants of indus-trial buyer behavior. Journal of Marketing, 53(April), 66-79.

Meyer, J. P., Allen, N. J., & Smith, C. A. (1993). Commitment to organizations and occupations:Extension and test of a three-component conceptualization. Journal of Applied Psychology, 78,538-551.

Morgan, R. M., & Hunt, S. D. (1994). The commitment-trust theory of relationship marketing.Journal of Marketing, 5S(July), 20-38.

Oliva, T. A., Oliver, R. L., & MacMillan, I. C. (1992). A catastrophe model for developing servicesatisfaction strategies. Journal of Marketing, 56, 83-95.

Park, S. H. (1996). Relationships between involvement and attitudinal loyalty constructs in adultfitness programs. Journal of Leisure Research, 28, 233-250.

Pedhazur, E. J., 8c Schmelkin, L. P. (1991). Measurement, design, and analysis: An integrated approach.Hillsdale, NJ: Lawrence Erlbaum Associates.

Piaget, J. (1970). The stages of the intellectual development of the child. In P. Mussen, J. Conger,& J. Kagan (Eds.), Readings in child development and personality (pp. 291-302). New York:Harper & Row.

Pritchard, M. P. (1992). Development of the psychological commitment instrument for measur-ing travel service loyalty (Doctoral dissertation, University of Oregon, 1991). DissertationAbstracts International, 52, 4408A.

Pritchard, M. P., & Howard, D. R. (1997). The loyal traveler: Examining a typology of servicepatronage. Journal of Travel Research, ^5(Spring), 2-10.

Pritchard, M. P., Howard, D. R., & Havitz, M. E. (1992). Loyal measurement: A critical exami-nation and theoretical extension. Leisure Sciences, 14, 155-164.

Pritchard, M. P., Havitz, M. E., & Howard, D. R. (1999). Analyzing the commitment-loyalty linkin service contents. Journal of the Academy of Marketing Science, 27(3), 333-348.

Ray, K. M. (1997). The effects of enduring involvement as a mediating variable between con-sumers' motivations and their leisure behavior. Dissertation Abstracts International Section A:Humanities and Social Sciences, 58(6-A), 2305.

Rosenberg, M. J. (1960). An analysis of affective-cognitive consistency. In M.J. Rosenberg, C. I.Hovland, W.J. McQuire, R. P. Abelson, &J. W. Brehm (Eds.), Attitude organization and change:An analysis of consistency among attitude components (pp. 15-64). New Haven CT: Yale UniversityPress.

Schuett, M. A. (1993). Refining measures of adventure recreation involvement. Leisure Sciences,15, 205-216.


Selin, S., Howard, D. R., Udd, E., & Cable, T. (1988). An analysis of consumer loyalty to munic-ipal recreation programs. Leisure Sciences, 10, 210-223.

Sem, J., & Vogt, C. A. (1997). Demarketing as a new communication tool for managing publicland use. Trends, 43(4), 21-25.

Siegenthaler, K. L., 8c Lam, T. C. M. (1992). Commitment and ego-involvement in recreationaltennis. Leisure Sciences, 14, 303-315.

Stebbins, R. A. (1998). After work: The search for an optimal leisure lifestyle. Calgary, AB: DetseligEnterprises Ltd.

Stebbins, R. A. (2000). Serious leisure. In E. L.Jackson & T. Burton (Eds.), Leisure studies: Prospectsfor the twenty-first century (pp. 69-80). State College, PA: Venture.

Warrington, P., & Shim, S. (2000). An empirical investigation of the relationship between prod-uct involvement and brand commitment. Psychology and Marketing, 17, 761-782.

Documents

Examining Relationships between Leisure Involvement ... · loyalty, stating that "the concept of commitment provides an essential basis for distinguishing between brand loyalty and